What hardware do you need to run DeepSeek V4-Flash?

DeepSeek V4-Flash at UD-Q4_K_M is 155 GB of weights, with a context ceiling of 1M tokens. The DeepSeek that replaced the R1 distills. Its compressed cache makes a million-token context practical. On the cheapest machine that runs it, a Mac Studio M5 Ultra, 256GB, it generates 49 tok/s at 32k of context, worked out from memory bandwidth.

How good is DeepSeek V4-Flash, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 35 (max effort; the 0731 build), which puts it in the Sonnet-class band. In the same league as the labs' mainstream models on this index. See the whole table.

Head to head with its neighbours on the leaderboard: vs Qwen3.8 Flash Next above it, vs Qwen3.8 27B below. Two more models need much the same memory at 32k of context: vs MiniMax M2.7, vs Inkling Small.

What DeepSeek V4-Flash costs either way

Renting the same model costs $0.04732 per million input tokens and $0.09464 per million output (OpenRouter, cheapest active endpoint at 15:1 input to output, checked 2026-10-29). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill. What a million tokens costs each way puts the two prices side by side.

Machines that run DeepSeek V4-Flash

The Mac Studio M5 Ultra, 256GB holds it to 256k, the longest the calculator offers.

The table is the cheapest machine in each family, not the only one that runs it: 5 of the 56 machines on this site hold it at 32k. Within a family only memory decides, so that is every Strix Halo box with 192GB or more and every Mac Studio with 256GB or more. One table sets every machine against the others: price, the memory its GPU can use, what it holds and how long it takes to pay for itself.

One machine per family, cheapest first. Speeds are measured where a public benchmark exists and estimated from memory bandwidth otherwise; the calculator says which for any configuration. The longest context is the longest setting the calculator offers that the machine still holds this model at, cache included, and no machine is shown taking it past its own 1M limit. Each one opens the calculator on that machine at that length.

For scale, the calculator starts from 80 tok/s for a hosted API and times a local machine against it. The one above does not reach it, at 49 tok/s.

The cheapest machine that runs it is sold with 256 GB. There is a page on what 256 GB runs, machine by machine: it names the 2 machines here sold at that size, and the 44 current models the roomiest of them holds at 32k, this one among them.

It fits in less memory than that: the smallest size here that runs it at 32k is 192 GB, on the Framework Desktop, 192GB, which hands a model 160 GB.

Run DeepSeek V4-Flash at 256k on the Mac Studio M5 Ultra, 256GB

How much memory does DeepSeek V4-Flash need at 256k?

Every figure above is priced at 32k of context. At 256k, the longest the calculator offers, DeepSeek V4-Flash needs 156.9 GB rather than 155.3 GB. The weights are the same either way; the key-value cache goes from 0.2 GB to 1.8 GB.

The window costs nothing in hardware: the one machine here still sold at a published price that holds DeepSeek V4-Flash at 32k, the Mac Studio M5 Ultra, 256GB at $10,799, holds it at 256k as well.

On the Mac Studio M5 Ultra, 256GB it runs 49 tok/s estimated at 32k and 40 tok/s estimated at 256k. Every token generated reads the whole cache back, so a longer window is slower as well as bigger.

The machines that miss DeepSeek V4-Flash, and what they run

Of the machines here with a published price that are still sold, every one but the Mac Studio M5 Ultra, 256GB misses DeepSeek V4-Flash, and none of them misses by a window: its 155 GB of weights are larger than the usable memory in every one of them, before a single token of key-value cache. The DGX Spark GB10 Grace Blackwell, 128GB comes nearest, 35.6 GB short at 4k, the shortest window on the list. Qwen3.8 27B is the strongest model seven of the eight hold, and it scores 34 on the intelligence index against this model's 35.

Four machines here do hold DeepSeek V4-Flash at 32k, and none of them is an answer to what to buy, which is the question the rest of this page asks: the Mac Studio M3 Ultra, 256GB and the Mac Studio M3 Ultra, 512GB, no longer sold, at the $7,099 and $9,499 they launched at, and the Strix Halo Framework Desktop, 192GB and the Mac Studio M5 Ultra, 512GB, on sale with no published price. They are sold with 192 GB, 256 GB and 512 GB of memory and hand a model 160 GB to 384 GB of it, against the 155 GB DeepSeek V4-Flash needs at 32k.

One machine per family, the largest configuration in each that still misses, nearest first. Short by is the weights plus the key-value cache at 4k measured against the memory that machine's GPU can address, which that machine's own page gives. The last column is the highest-scoring current model the machine holds at 32k.

Price the DGX Spark GB10 Grace Blackwell, 128GB on Qwen3.8 27B instead

The specifics

Parameters

284B, of which 13B are active per token

Quantisation

UD-Q4_K_M

Weights on disk

155 GB

KV cache

0.2 GB at 32k context — Every layer compresses its cache by a different ratio: 21 layers keep one entry per four tokens, 20 keep one per 128, and two are a fixed 128-entry window. Summed, that is about 6.9 KB a token for the whole model — roughly a thirteenth of what the head counts imply. Read from the repo's own inference code. How weights and cache add up.

Maximum context

1M tokens

Licence

MIT

Sources

deepseek-ai/DeepSeek-V4-Flash-0731